AComprehensive Cloud Framework for Scalable, Reliable, and Replicable Human Connectome Es- timation and Meganalysis

The expansion of publicly available multimodal MR datasets enables analysis of the structure and function of the human brain at an unprecendented scale. Alongside development of tools and standards for this data, neuroinformatics studies regularly uncover evidence for patterns between behaviour and structure of the brain. As the field of data collection paradism and processing tools grows, findings made across studies are becoming increasingly di cult to compare. We have developed a turn-key pipeline for reliable structural connectome estimation at scale, and use it as a framework for harmonized data processing and performing “meganalaysis” across collections of data. We demonstrate this framework by estimating 2,861 connectomes, and by virtue of harmonized data processing show that while we can yield qualitatively similar graphs, results are meaningfully quantitatively di erent across datasets.